| | Using Sibling Differences to Estimate Effects of Parenting on Adolescent Sexual Risk BehaviorsReceived 9 September 2007; accepted 21 December 2007. published online 25 April 2008. Abstract PurposeTo estimate effects of positive and involved parenting during mid-adolescence on sexual risk behaviors (frequency of intercourse, unprotected intercourse, and number of sexual partners) during late adolescence. Substantial literature suggests that supportive family contexts and parenting behaviors may discourage adolescents from engaging in early and risky sexual activities; yet methodological limitations hamper the conclusions regarding causality and directionality that can be drawn from much existing research. To address such limitations, the current study used a variety of increasingly conservative statistical modeling techniques to help control for unobserved heterogeneity and potential bias and hence to progress toward identifying causal relationships. MethodsDrawing from a nationally representative longitudinal survey of adolescents (NLSY97; N = 4980), this study used ordinary least squares (OLS) regression models, lagged regression models, and family fixed-effects models to assess whether parental knowledge, parent negativity, and family activities during midadolescence predicted differences in late adolescent sexual risk behaviors. ResultsEven after controlling for unobserved heterogeneity across individuals and across families, parenting processes significantly predicted later adolescent sexual risk behaviors. Specifically, more regular family activities and less negative and hostile parenting during mid-adolescence predicted lower sexual risk behaviors during late adolescence. ConclusionsResults concerning the buffering effects of parenting on adolescent risk behaviors help to inform prevention and intervention efforts. Through the use of more rigorous statistical methodology and large representative samples of youth, this research provides an exemplar of how survey research can seek to move closer to understanding causal processes in the exceedingly complex systems of human development. Sexual activity has become normative for adolescents in the United States, with more than two-thirds of adolescents engaging in sexual intercourse before age 19 years [1]. Early sexual activity nonetheless carries substantial health and psychosocial risks. In particular, the frequency of sexual activity, lack of reliable birth control, and the number of sexual partners substantially increase the risk of negative outcomes such as sexually transmitted diseases and pregnancy for adolescents [2]. Hence a broad and comprehensive focus on factors that affect adolescents' sexual risk behaviors is important to further understanding of youth health risks in the United States [3]. Conceptual and empirical literature asserts that family processes and parenting behaviors are important contributors to adolescent sexual behaviors, aiding the development of emotional and decision-making skills and protecting adolescents from negative peer influences [4], [5], [6]. Past studies provide important initial insights and have helped to identify particular parenting processes that may be protective against risky sexual activity, most notably parental monitoring or knowledge of adolescent activities and friends, regular family activities and routines, and low parental negativity and psychological control [4], [6]. However the empirical research base on whether parenting processes actually affect adolescent sexual risk behaviors is somewhat equivocal. The current research seeks to address multiple limitations of previous research [4]. First, much research has incorporated unidimensional, dichotomous measures of sexual experience, rather than more comprehensive measures of sexual risk behaviors [7], [8]. Numerous studies have used cross-sectional data from one time [9], leaving unaddressed the thorny issues of temporality and directionality. Other studies provide longitudinal data and temporal ordering such that family processes at one point predict sexual risk behaviors at a later point [8], [10]. Even in longitudinal models, however, omitted variable bias remains a concern. Differences across adolescents (e.g., antisocial tendencies), across families (e.g., high-risk neighborhoods), or across siblings within families (e.g., different parental expectations of older vs. younger siblings) all may be linked both with less effective parenting processes and with adolescent risk behaviors. If unmeasured, such heterogeneity would lead to correlations but not causal links between parenting processes and adolescent sexual behaviors. The goal of this study was to use increasingly conservative statistical modeling techniques to help control for unobserved heterogeneity and potential bias and hence progress toward more causal analyses assessing whether positive and involved parenting helps protect adolescents from risky sexual activity. We hypothesized that greater parental knowledge, less negativity, and more regular family activities would predict lower levels of adolescent sexual risk behaviors. Methods  Sampling and data collection Data for this investigation were drawn from the National Longitudinal Study of Youth 1997 (NLSY97), an annual survey study of youth in America born between 1980 and 1984. The original sample consisted of a nationally representative group of 8984 adolescents between the ages of 12 and 16 years drawn from 6819 unique households, with purposive oversampling of economically poor and minority youth. Adolescents have been interviewed annually since 1997, with wave 1 parent interviews and annual income updates. The sample has had remarkably low attrition, with 93% of respondents followed in wave 2 and nearly 88% followed into wave 6. Analyses focus on two samples from the NLSY97: a total sample of all youth aged 12–14 years at screening (n = 4980) and a subsample of siblings (n = 1058). Excluded were respondents >14 years of age at screening (n = 3567) who did not report on parenting, and youth without valid data on family processes or sexual behaviors (n = 437). Related to their older age, excluded respondents came from families with fewer minors, more income, higher parental education, and lower parental employment. Analyses draw data from the first six waves, split into two time periods: Time 1 includes waves 1–3, which cover the ages 12–14, 13–15, or 14–16, depending on adolescents' age at entry (average age, 14 years). We term Time 1 “mid-adolescence.” Time 2 includes waves 4–6, covering the years 15–17, 16–18, or 17–19 (average age, 17 years), which we term “late adolescence.” Measures Family process variables Adolescent perceptions of three dimensions of family processes were assessed in waves 1, 2, and 3, all derived from youth report and pertaining to residential parents: parental negativity, parental knowledge, and family activities. Parental negativity was assessed with an average of three questions concerning each parent's negative, psychologically controlling behaviors (How often does s/he… criticize you or your ideas? blame you for her problems? make plans with you and cancel for no good reason?; 0 = never, 4 = always; α =.70–.72). Parental knowledge was measured with an average of four items for each parent (How much does s/he know about … whom you are with when you are not at home? your close friends? your close friends' parents? whom your teachers are and what you are doing in school?; 0 = knows nothing, 4 = knows everything; α = .85–.87). Family activities was an average of three items (Number of days per week you … typically eat dinner with family? typically have fun with your family? family does something religious? [0 = no days per week, 7 = 7 days per week]). In two-parent families mother and father subscales on knowledge and negativity were averaged. Each of the three family process measures were then averaged across waves 1–3 to increase measurement reliability, decrease measurement error, and capture adolescents' experiences over mid-adolescence [11]. Risky sexual behavior Youth reported on their sexual behaviors at age 14 and annually thereafter. Three items assessed number of sex partners, frequency of sexual intercourse, and frequency of unprotected sexual intercourse (defined as using a method other than a condom, diaphragm, birth control pills, intrauterine device, or injectable agent) since the last interview. Because items were measured in different units, they were first standardized into deciles and then averaged (α = .72–.82) into a risky sexual behavior composite. Scores were averaged for waves 1–3 for a measure of mid-adolescent risky sexual behavior (adolescents were aged 14–16 for these reports). Scores from waves 4-6 (adolescents aged 15–19) were averaged for late-adolescent risky sexual behavior, the outcome variable of interest in the analyses. Demographic characteristics Important demographic characteristics of youth and their families were also included in analyses, including minors in the household, parent employment, family structure, adolescent age (to control for entering the sample at different ages), ethnicity, and gender, as reported by youth; and parent age, parent education, and parent income, as reported by parents. Imputation of missing demographic control variables was conducted using a maximum likelihood approach with expectation maximization [12]. Data analysis Data analyses used a series of increasingly conservative statistical modeling strategies to control for omitted variable bias and attempt to move toward identifying more causal relationships between parenting processes and adolescent sexual risk behaviors. All models used family processes and control variables at Time 1 during mid-adolescence (waves 1–3) to predict sexual risk behaviors at Time 2 during late adolescence (waves 4–6), thus providing temporality. Model 1 uses a standard longitudinal OLS regression model: where Sex at Time 2 is a measure of risky sexual behavior for individual i, Parenting at Time 1 represents a vector of parenting processes for individual i, and Controls at Time 1 represents a vector of control variables for individual i. Model 2 adds the Time 1 risky sex composite as a predictor. This lagged (residualized or autoregressive) regression model thus controls for unmeasured factors that vary across individuals and have an invariant effect on sexual behaviors [13], [14], and helps control for potential bidirectionality, or child effects on parenting during mid-adolescence [15]. Model 3 incorporates a family fixed-effect model (also called a sibling-difference model), in which measured and unmeasured characteristics that are invariant within families are differenced out of the equation [11], [16]. Here the family average for sex at Time 2 for each family, f, is differenced from the Time 2 sex score for each individual, i, in family f. Similar differencing occurs for each parenting measure and control measure on the right side of the equation. This differencing controls for unobserved characteristics that vary across families and have consistent effects within families on the outcome of interest (e.g., neighborhood contexts or shared genetic influences). Family fixed-effects models require heterogeneity within families on both the central predictors and dependent variables of interest, in this case parenting processes and adolescent sexual risk behaviors. Developmental and genetics research provides evidence toward these claims [17], [18]. Model 3 may still be biased if unmeasured differences have a differential influence on adolescent sexual risk behaviors within families, for instance, if one sibling experienced early puberty. Sibling models also require heterogeneity in the predictors to be exogenous. If parents treat children in the same family differently as a response to early sexual behavior, this assumption would be threatened. To address such biases, model 4 incorporates a lag, adding differences in Time 1 sexual risk behaviors as a predictor, thus controlling for both measured and unmeasured heterogeneity that is linked with consistency in sexual risk behaviors within individuals across time, and across families. Results Longitudinal OLS regression results Longitudinal regression models were used for both the total sample of adolescents (Table 2) and the sibling subsample (Table 3). Because results were similar, only results for the sibling subsample are discussed in detail. Model 1 used a standard longitudinal ordinary least squares (OLS) regression model with robust standard errors and clustering to allow the errors to be correlated within families. Across all models, youth age was a strong predictor of risky sexual activity, indicating the normative growth of sexual behavior through adolescence. Controlling for differences in age as well as other demographic and socioeconomic characteristics of adolescents and their families, it was found that youth who reported more regular family activities, higher parental knowledge, and lower negative parenting during mid-adolescence experienced significantly lower sexual risk behaviors during later adolescence. To assess the size of the effects, marginal effects (elasticities) are shown, which indicate the percent change in the dependent variable for each 1% change in the predictor, holding other variables constant at their sample mean. A 1% increment in family activities, parental knowledge, and parent negativity predicted .33%, .23%, and .14% differences, respectively, in sexual risk behaviors, which are small to moderate- sized effects. Model 2 presents the lagged OLS regression model. The coefficient for earlier risky sexual activity is highly significant, indicating the stability of risky sexual practices through adolescence; however parenting processes still largely retained importance. Coefficients and marginal effects for family activities and negativity declined but maintained significant links with sexual risk behaviors in late adolescence. The coefficient for parental knowledge dropped more precipitously, becoming nonsignificant and suggesting that once earlier risk behaviors were accounted for, parental knowledge did not predict a significant change in adolescent sexual risk behaviors. A 1% increment in family activities, parental knowledge, and parent–child negativity predicted change in sexual risk behaviors of .22%, .04%, and .11%, respectively, which were small effects. Overall the results suggest that unmeasured factors that had a consistent effect on adolescent risky sexual activity were also related to parenting, inflating links between parenting and risky sexual activity in model 1. Controlling for these factors, more regular family activities and less parental negativity predicted lower growth in adolescent risky sexual activities. Family fixed-effects regression results Models 3 and 4, presented in Table 4, used the family fixed-effects specification. Age was a strong predictor, indicating that within-family differences in sexual risk behaviors were strongly related to sibling age differences, again reflecting normative growth of sexual behaviors over adolescence. Coefficients for the parenting variables were smaller and standard errors larger than in the OLS sibling sample model, reflecting lower precision. However family activities still retained statistical significance, with parent-child negativity at trend level. The marginal effects indicate that a sibling who reports 1% higher family activities than his or her family average in turn experiences .26% lower sexual risk behaviors than the family average. A 1% difference in negativity predicted .10% higher sexual risk behaviors. Parental knowledge was not significant, although the marginal effect was larger than negativity at .14%. Finally, model 4 presents results from the family fixed-effects model, controlling for sibling differences in earlier risky sexual activity. Results are quite similar to those of model 3. Family activities and negativity significantly predicted later adolescent sexual risk behaviors, with elasticities of .23% and .11%. Parental knowledge, albeit not significant, showed an effect size similar to that of negativity (.12%). This very conservative estimation suggests that after controlling for sibling differences in age, earlier sexual risk behaviors, and other characteristics, youth who experienced more positive and involved family processes than their siblings during mid-adolescence in turn showed lower growth in risky sexual behaviors than their siblings through late adolescence. In summary, the OLS and fixed-effects models produced very similar patterns of findings, with the fixed-effects models providing coefficients that were more consistent and less biased by unobserved heterogeneity. Discussion Previous research has revealed significant relationships between parenting practices and adolescents' engagement in risky sexual activities. Yet a host of methodological limitations have curtailed researchers' ability to assess causal links and to rule out the possibility that unmeasured heterogeneity, or bias, drives correlations between more positive and involved parenting practices and lower adolescent sexual risk behaviors [4]. The analyses presented here assessed a nationally representative sample of American adolescents followed for 6 years, using modeling strategies designed to control for unmeasured heterogeneity to delineate a clearer picture of whether parenting practices help to protect adolescents from sexual risk. Replicated over increasingly conservative models, our results provide strong evidence that experiencing less negative parenting behaviors and engaging in more regular activities with family during mid-adolescence helps youth to avert risky sexual behaviors in late adolescence, whereas less consistent evidence emerged regarding protective effects of parental knowledge. Theoretical contributions The current study extends previous research [3], [19] on the relationship between negative parenting and sexual risk behaviors. Our results found that more negative and psychologically controlling parenting during mid-adolescence predicted higher levels of sexual risk behaviors (such as multiple partners, more frequent intercourse, and lack of effective birth control) during late adolescence. Negative and psychologically controlling parenting behaviors may inhibit adolescents' development of self-efficacy and identity, interfere with mature and responsible decision-making skills, and affect the development of healthy relationships, in turn leading to an elevated likelihood of engaging in risky behaviors [3], [5], [20]. These results extend earlier research that has primarily linked psychologically controlling parenting to internalizing problems such as depression [20]. In addition to the emotional tenor of parenting behaviors, adolescents' engagement in regular activities with their families was also protective. Family activities or rituals have recently been identified as centrally important supports for children, providing opportunities for emotional warmth, communication, and transmission of values and beliefs [21]. Yet limited empirical evidence to date has linked family activities with youth development [8], [22], [23]. One recent paper found that more regular family activities during mid-adolescence predicted a lower likelihood of having engaged in sexual intercourse by late adolescence in a sample of Scottish youth [8]. The current results extend this sparse research base, suggesting that even within families, differences in adolescents' participation in family activities were prospectively linked to differences in risky sexual behaviors. Further research should seek to delineate the mechanisms through which engagement in such activities is protective for youth. Finally, results concerning parental knowledge of adolescents' peers and teachers, a central component of parental monitoring and behavioral control, showed mixed findings. In the less stringent models, greater parental knowledge during mid-adolescence predicted lower adolescent sexual risk behaviors during late adolescence. This finding was not robust across model specifications, however, and once unobserved heterogeneity across individuals and families was controlled for, the relationship became nonsignificant (albeit similar in size to that of parental negativity). Some have argued that parental knowledge is more a reflection of adolescent disclosure and hence influenced by adolescent behaviors, rather than being an assessment of parents' monitoring and behavioral control efforts that affect adolescent behaviors [24], [25]. Others have argued for a bidirectional model in which processes are driving both directions [22]. Results from the current analyses cannot definitively parse these different possibilities, but suggest that unmeasured heterogeneity may be important. Another possible explanation for our pattern of results is that parental knowledge provides short-term benefits for adolescents [22], for instance by limiting their access to risky situations and peers. However if these experiences do not help youth to develop self-regulatory behaviors and decision-making skills or to internalize parental beliefs, they may not confer long-term benefits when youth gain more freedom and experience less parental oversight in later adolescence. Methodological contributions and limitations In addition to the theoretical goals, this research also sought to contribute methodologically by assessing a series of increasingly conservative regression models to systematically control for measured and unmeasured heterogeneity across individuals and families in estimating links between mid-adolescent family processes and late adolescent risky sexual behaviors. If unmeasured heterogeneity (i.e., unobserved factors that affect the central construct of interest, in this case adolescent risky sexual activity) are also related to the central predictors of interest (here parenting processes), then the estimates of parenting will be biased. Analytic models moved from longitudinal OLS regressions, to lagged OLS regressions, to longitudinal and lagged family fixed-effects models, all controlling for age-related differences in sexual activity as well as other observed differences in adolescent and family demographic characteristics. Family fixed-effects models exploit measured heterogeneity within families while controlling for unmeasured heterogeneity across families which affects all siblings in a family similarly. Finally, the lagged family fixed-effects models controlled for unobserved factors that have a consistent effect on individuals' risky sexual activity, and also controlled for possible child elicitation or bidirectionality (i.e., that parenting during mild-adolescence was driven in part by adolescent's early sexual risk behaviors). Lagged fixed-effects models present the strongest controls for unmeasured heterogenteity and hence provide conservative estimation of links between parenting and adolescent sexual behaviors. Even with such controls, results unearthed significant protective effects for family activities and low parental negativity. It is important, however, to acknowledge the limitations of fixed-effects models. Fixed-effects models produce less precise estimates, biasing results toward nonsignificance [16]. Another concern is potential bias introduced through constraints on the sample. By including only those individuals in families with more than one child, the sample becomes less representative and generalizable, potentially over-representing larger families or families with particular unobserved characteristics [26]. As Moffitt [27] notes, such models prioritize internal validity while sacrificing external validity, or generalizability. Moreover fixed-effects models cannot control for all unobserved characteristics that vary across individuals, although in the lagged fixed-effects models such factors would bias the findings only if they had differential effects across siblings and across time on sexual risk behaviors. Beyond the statistical limitations, other study limitations deserve mention. First, data were nearly all drawn from one reporter, the adolescent. As such, reports of parenting behaviors reflect adolescents' perceptions of their family environments; other reporters might present a different view. The use of a single reporter also raises concerns over shared method variance, here attenuated through the use of lagged and fixed-effect methodology. Finally, although youth age and earlier risky sexual activities were both strong predictors of later risky sexual activity, these controls likely did not capture all unmeasured heterogeneity, in part because of limited variability in sexual activity during mid-adolescence. Further study should more explicitly assess entrance into sexual activity and growth in risky sexual behaviors, perhaps through trajectory analysis. Conclusions In summary, results from this study suggest that parenting and family processes may serve as important influences on adolescent risky sexual behaviors. Regular family activities may buffer adolescents, whereas negative parenting behaviors may increase adolescents' engagement in sexual risk behaviors. Given the substantial potential health and psychosocial risks of multiple sexual partners and lack of effective contraception, these findings have important implications for public health and intervention/prevention efforts. Through the use of more rigorous statistical methodology and large, representative samples of youth, this research provides an exemplar of how survey research can seek to move closer to understanding causal processes in the exceedingly complex systems of human development. Acknowledgments This research was funded by Grant 2538 from the W. T. Grant Foundation; by Grant 1R03HD055229-01A1 from the National Institutes of Health; by the generous support of a Fulbright Senior Scholar Award from the Australian-American Fulbright Commission; and by a Senior Visiting Fellow position at the Social Policy Research Centre, University of New South Wales, Australia (to R.L.C.). References [1]. [1]Terry-Humen E, Manlove J, Cottingham S. Trends and recent estimates: Sexual activity among US teens. Washington, DC: Child Trends; 2006;. [2]. [2]Franzetta K, Ikramullah E, Manlove J, et al. Facts at a glance. Washington, DC: Child Trends; 2006;. [3]. [3]Rodgers K. Parenting processes related to sexual risk-taking behaviors of adolescent males and females. J Marriage Fam. 1999;61:99–109. [4]. [4]Buhi ER, Goodson P. Predictors of adolescent sexual behavior and intention: A theory-guided systematic review. 2007;40:4–21. [5]. [5]Gray MR, Steinberg L. Unpacking authoritative parenting: Reassessing a multidimensional construct. J Marriage Family. 1999;61:574–587. [6]. [6]Miller BC, Benson B, Galbraith KA. Family relationships and adolescent pregnancy risk: A research synthesis. Develop Rev. 2001;21:1–38. [7]. [7]Ream GL, Savin-Williams RC. Reciprocal associations between adolescent sexual activity and quality of youth–parent interactions. J Fam Psychol. 2005;19:171–179. MEDLINE |
CrossRef
[8]. [8]Sweeting H, West P, Richards M. Teenage family life, lifestyles and life changes: Associations with family structure, conflict with parents and joint family activity. Int J Law Policy Fam. 1998;12:15–46. [9]. [9]Huebner AJ, Howell LM. Examining the relationship between adolescent sexual risk-taking and perceptions of monitoring, communication, and parenting styles. J Adolesc Health. 2003;33:71–78. |
CrossRef
[10]. [10]Wight D, Williamson L, Henderson M. Parental influences on young people's sexual behavior: A longitudinal analysis. J Adolesc. 2006;29:473–494. MEDLINE |
CrossRef
[11]. [11]Aaronson D. Using sibling data to estimate the impact of neighborhoods on children's educational outcomes. J Hum Resources. 1998;33:915–946. [12]. [12]Dempster AP, Laird NM, Rubin DB. Maximum likelihood from incomplete data via the EM algorithm. J R Stat Soc. 1977;39:1–38. [13]. [13]Cain GG. Regression and selection models to improve nonexperimental comparisons. In: Bernett CA, Lumsdiane AA editor. Evaluation and Experiment. New York: Academic Press; 1975;p. 297–317. [14]. [14]Kessler R, Greenberg EF. Linear Panel Analysis: Models of Quantitative Change. New York: Academic Press; 1981;. [15]. [15]Bell RQ. A reinterpretation of the direction of effects in studies of socialization. Psychol Rev. 1968;75:81–95.
CrossRef
[16]. [16]Duncan GJ, Magnuson KA, Ludwig J. The endogeneity problem in developmental studies. Res Hum Dev. 2004;1:59–80. [17]. [17]Deater-Deckard K, Fulker DW, Plomoin R. A genetic study of the family environment in the transition to early adolescence. J Child Psychol Psychiatry. 1999;40:769–775. MEDLINE |
CrossRef
[18]. [18]Harris KM, Morgan SP. Fathers, sons, and daughters: Differential paternal involvement in parenting. J Marriage Fam. 1991;53:531–544. [19]. [19]Luster T, Small SA. Factors associated with sexual risk-taking behaviors among adolescents. J Marriage Fam. 1994;56:622–632. [20]. [20]Barber BK. Parental psychological control: Revisiting a neglected construct. Child Dev. 1996;67:3296–3319. MEDLINE |
CrossRef
[21]. [21]Fiese BH, Tomcho TJ, Douglas M, et al. A review of 50 years of research on naturally occurring family routines and rituals: Cause of celebration?. J Fam Psychol. 2002;16:381–390. MEDLINE |
CrossRef
[22]. [22]Coley RL, Votruba-Drzal E, Schindler H. Trajectories of parenting processes and adolescent substance use: Reciprocal effects. J Abnorm Child Psychol. 2008;36:613–635.
CrossRef
[23]. [23]Eisenberg ME, Olson RE, Neumark-Sztainer D, et al. Correlations between family meals and psychosocial well-being among adolescents. Arch Pediatr Adolesc Med. 2004;158:792–796. MEDLINE |
CrossRef
[24]. [24]Kerr M, Stattin M. What parents know, how they know it, and several forms of adolescent adjustment: Further support for a reinterpretation of monitoring. Dev Psychol. 2000;38:366–380. [25]. [25]Kerr M, Stattin H. Parenting of adolescents: Action or reaction?. In: Crouter AC, Booth A editor. Children's Influence on Family Dynamics: The Neglected Side of Family Relationships. Mahwah, NJ: Lawrence Erlbaum Associates; 2003;p. 121–151. [26]. [26]Vartanian TP, Buck PW. Childhood and adolescent neighborhood effects on adult income: Using siblings to examine differences in ordinary least squares and fixed-effects models. Soc Serv Rev. 2005;29:60–94. [27]. [27]Moffitt R. Remarks on the analysis of causal relationships in population research. Demography. 2005;42:91–108. MEDLINE |
CrossRef
a Social Policy Research Centre, University of New South Wales, New South Wales, Australia b Applied Developmental and Educational Psychology, Boston College, Chestnut Hill, Massachusetts c Jewish Family and Children's Service, Boston, Massachusetts d Center on the Developing Child, Harvard University, Cambridge, Massachusetts  Address correspondence to: Rebekah Levine Coley, Ph.D., Applied Developmental & Educational Psychology, Boston College, Campion Hall 239A, 140 Commonwealth Avenue, Chestnut Hill, MA 02467.
PII: S1054-139X(08)00101-8 doi:10.1016/j.jadohealth.2007.12.012 © 2008 Society for Adolescent Medicine. Published by Elsevier Inc. All rights reserved. | |
|
FULL TEXT
